Genetic Algorithm Based Feature Selection to Optimize Cardiovascular Disease Prediction Using Classification Techniques

Abstract

Background: A correct assessment of cardiac illness has the potential to save a person's life, while a poor prognosis can be fatal. Still, there are cases that are diagnosed wrongly, prognoses, then cured. People are encouraged to participate in diverse diagnostic examinations. Such diagnostics are frequently ineffective at detecting problems at this time. The purpose of the work is to find a way that effectively anticipates the incidence of cardiovascular problems using fewer variables while sparing patients' time and money spent on diagnostic procedures. The dataset, which included 14 parameters, were gathered from digital sources for this research Work. A patient undergoes fewer tests by using the genetic algorithm to identify the qualities that are most helpful in the diagnosis of heart issues. A genetic search decreased 14 traits to 6 attributes. In order to predict disease with more accuracy than before the reduction of data, four classification techniques Random Forest (RF), AdaBoost, k-Nearest Neighbor (k-NN), and Support Vector Machine (SVM) applied. The accuracy values of the Random Forest system, the AdaBoost Method, the k-NN template, the SVM framework, and the Model of Random Forest were 92.90%, 89.30%, 90.10%, and 93.70%, respectively. Following the application of the genetic algorithm, the Random Forest model's accuracy was 96.90%, followed by the AdaBoost method's accuracy of 92.30%, the k-NN algorithm's accuracy of 94.10, and the SVM model's accuracy of 98.90%. In comparison to the Random Forest (RF), AdaBoost, and k-NN classifiers, after attribute selection using genetic algorithms, the SVM classifier delivers excellent results and high metric values for heart disease prediction.